Gabor Wavelet Feature Based Face Recognition Using the Fractional Power Polynomial Kernel Fisher Discriminant Model

This paper presents a technique for face recognition that uses Gabor wavelets with five scales and eight orientations to derive desirable facial features characterized by spatial locality, spatial frequency, and orientation selectivity to cope with the variations due to illumination and facial expression changes. The fractional power polynomial kernel principal component analysis (KPCA) method maps the input data into an implicit feature space with a non linear mapping. Being linear in the feature space, but nonlinear in the input space, KPCA is capable of deriving low dimensional features that incorporate higher order statistic. The Fisher classifier is applied to Gabor featured KPCA mapped data. The effectiveness of this Gabor kernel Fisher classifier (GKFC) algorithm is compared with some of the existing algorithms for face recognition using the FERET and the ORL databases. This algorithm performs better than some of the existing popular algorithms.